Systems and Soft Computing最新文献

{"title":"Sentiment analysis of movie reviews: A flask application using CNN with RoBERTa embeddings","authors":"Biplov Paneru ,&nbsp;Bipul Thapa ,&nbsp;Bishwash Paneru","doi":"10.1016/j.sasc.2025.200192","DOIUrl":"10.1016/j.sasc.2025.200192","url":null,"abstract":"<div><div>Sentiment analysis, an important task in Natural Language Processing (NLP), focuses on identifying and extracting sentiments from input. With the exponential expansion of digital information, sentiment analysis has recently gained significant attention across various domains. Traditional sentiment analysis methods paired with static embeddings often fall short in capturing the deep contextual relationships within text. In this work, we analyze sentiment in IMDB movie reviews using a hybrid deep learning model combining RoBERTa embeddings with a convolutional neural network (R-CNN). We provide a comprehensive overview of the creation and assessment of a convolutional learning model especially suited for sentiment analysis of movie reviews using a dataset of around 50k entries. The proposed approach preprocesses movie reviews, employs RoBERTa to generate rich contextual embeddings, and processes these embeddings through a simple yet effective R-CNN architecture. We perform comprehensive analysis of the R-CNN model, showing a superior test accuracy of 91.5 %, achieving the best results compared to the baseline. Additionally, we develop a Flask-based application, demonstrating the practical applicability of our R-CNN model for real-time sentiment prediction.</div></div>","PeriodicalId":101205,"journal":{"name":"Systems and Soft Computing","volume":"7 ","pages":"Article 200192"},"PeriodicalIF":0.0,"publicationDate":"2025-01-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143149507","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Research and application of visual synchronous positioning and mapping technology assisted by ultra wideband positioning technology","authors":"Yiran Zhang,&nbsp;Lina Dong","doi":"10.1016/j.sasc.2025.200187","DOIUrl":"10.1016/j.sasc.2025.200187","url":null,"abstract":"<div><div>With the development of the intelligent era, improving the positioning accuracy and operational stability of robots has become an urgent problem that needs to be solved. This study combines the advantages and disadvantages of visual synchronous positioning and mapping technology, inertial measurement units, and ultra-wideband technology to design a combined positioning system. The system first uses the pre-integration method of the inertial measurement unit to align the inertial measurement unit with the camera frequency. Then, it uses a tightly coupled method to fuse the measurement data of the system and the inertial measurement unit, forming a visual-inertial system. The study uses extended Kalman filtering to fuse the constructed visual-inertial system with ultra-wideband technology, creating an ultra-wideband/visual-inertial integrated system. Finally, simulation analysis was conducted on the constructed composite system. The results indicated that the RMSE of the ultra-wideband/visual-inertial system under light and dark conditions were 0.0123 and 0.0212, and 0.0114 and 0.0123, respectively, in the motion trajectories with and without forming a loop. In extremely complex motion trajectories, the RMSE error of the research system was 0.0123. This indicates that regardless of the conditions, the research system has long-term robustness and high-precision positioning performance.</div></div>","PeriodicalId":101205,"journal":{"name":"Systems and Soft Computing","volume":"7 ","pages":"Article 200187"},"PeriodicalIF":0.0,"publicationDate":"2025-01-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143149506","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A dynamic financial risk prediction system for enterprises based on gradient boosting decision tree algorithm","authors":"Lin Ji ,&nbsp;Shenglu Li","doi":"10.1016/j.sasc.2025.200189","DOIUrl":"10.1016/j.sasc.2025.200189","url":null,"abstract":"<div><div>As financial technology develops, the dynamic prediction of enterprise financial risks has become a focus of attention in the financial field. The research aims to construct a dynamic financial risk prediction system for enterprises based on gradient boosting decision trees to improve the predicting accuracy and adaptability. The minimum absolute value shrinkage and selection operator algorithm were used for dynamic indicator selection. A dynamic prediction model was constructed by combining gradient boosting decision trees. The decision tree model parameters were optimized through gradient optimization using the sparrow search algorithm. The integrated model performed excellently on multiple evaluation indicators, with an area under the receiver operating characteristic curve of 0.8. The average accuracy was 92.38%, the recall was 94.27%, and the root mean square error and average absolute error were lower than other models, demonstrating high prediction accuracy and reliability. The average user satisfaction of this integrated model was 85%, significantly higher than the 46% of the ordinary gradient boosting decision tree model. This model can not only accurately identify risk situations, but also meet the actual needs of enterprise users. This study provides a new financial risk assessment tool for enterprises. This helps enterprises to identify and manage potential risks in a timely manner, which is of great significance for promoting healthy and sustainable development of enterprises.</div></div>","PeriodicalId":101205,"journal":{"name":"Systems and Soft Computing","volume":"7 ","pages":"Article 200189"},"PeriodicalIF":0.0,"publicationDate":"2025-01-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143149508","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Real-time facial reconstruction and expression replacement based on neural radiation field","authors":"Shenning Zhang ,&nbsp;Hui Li ,&nbsp;Xuefeng Tian","doi":"10.1016/j.sasc.2025.200185","DOIUrl":"10.1016/j.sasc.2025.200185","url":null,"abstract":"<div><div>It is now possible to do high-fidelity 3D facial reconstruction and unique view synthesis thanks to the recent discovery of Neural Radiance Fields (NeRF), which has established its substantial importance in the field of 3D vision. However, the operational approaches that are now in use require a significant amount of human engagement, such as the need for users to provide semantic masks and the inconvenience of manual attribute searching for non-expert users. Our approach focuses on enabling the manipulation of NeRF-reconstructed faces with just a single text input. A scene manipulator, specifically a conditional version NeRF with deformable latent codes, is the first thing that this paper trains to accomplish this objective, in dynamic scenes, allowing facial deformations to be controlled through latent codes. However, to synthesize local deformations in a variety of contexts, it is not desirable to describe scene deformations using only a single latent coding. Therefore, this paper proposes a text-driven operation pipeline for facial reconstruction with NeRF, the development of an operating network that is capable of learning to represent scene changes using latent codes that vary at different spatial locations, and the integration of a WeChat mini-program to facilitate practical applications. This application approach enables even non-expert users to easily synthesize novel views. Our method has achieved a certain breakthrough in the field of 3D facial reconstruction, providing users with a simple and convenient text-driven operation approach.</div></div>","PeriodicalId":101205,"journal":{"name":"Systems and Soft Computing","volume":"7 ","pages":"Article 200185"},"PeriodicalIF":0.0,"publicationDate":"2025-01-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143148526","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Local image style transfer algorithm for personalized clothing customization design","authors":"Xuemeng Wu","doi":"10.1016/j.sasc.2025.200183","DOIUrl":"10.1016/j.sasc.2025.200183","url":null,"abstract":"<div><div>With the increasing demand for personalized clothing from consumers, the style transfer technology of clothing images has become a key link in clothing customization design. However, when transferring clothing image styles in complex backgrounds, problems such as poor local image style transfer and boundary artifacts often arise. To address these issues, an attention mechanism-based approach to local style transfer in cyclic generative adversarial networks has been proposed. By introducing attention mechanisms, more precise probability allocation has been achieved. In addition, this study designs a local artifact correction model based on an improved residual network. The experimental results showed that the proposed method had an average ratio of 0.83 for the best performance image in user perception evaluation, which was at least 23.9 % higher than other methods. In addition, the average distance similarity of this research method reached 0.244, which was at least 4.4 % higher than other methods. In terms of mean square error, the research method had a mean square error as low as 3426, which was at least 8.5 % lower than other algorithms. In addition, regarding the artifact correction part, the average opinion score of the proposed method was 2.9, which was at least 7.4 % higher than other algorithms. The mean square error of this algorithm was only 16.13, at least 34.2 % lower than other algorithms. This study verifies the effectiveness of the proposed method in local style transfer and artifact correction of clothing images, provides strong technical support for the field of clothing customization, and helps to promote technical progress in this field.</div></div>","PeriodicalId":101205,"journal":{"name":"Systems and Soft Computing","volume":"7 ","pages":"Article 200183"},"PeriodicalIF":0.0,"publicationDate":"2025-01-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143148529","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Application of hybrid ant colony algorithm to the design of analog wavelet optimized circuits","authors":"Quanhui Ren,&nbsp;Chenyu Meng","doi":"10.1016/j.sasc.2025.200184","DOIUrl":"10.1016/j.sasc.2025.200184","url":null,"abstract":"<div><div>Facing the real-time and low-power requirements in non-smooth signal processing, the traditional digital wavelet transform method limits its efficiency and feasibility in practical applications due to the large amount of arithmetic and the need for A/D conversion. In order to overcome these shortcomings, the study proposes an analog circuit design method for rational approximation of wavelet function using hybrid ant colony algorithm. The study performs constrained mathematical modeling of the wavelet approximation through the minimum mean square error criterion and optimizes it using the hybrid ant colony algorithm. Also, the study designs a current-mode circuit based on the operational transconductance amplifier and current controlled conveyor second generation for implementing the analog wavelet transform. The results revealed that the amplitude response of the hybrid ant colony algorithm optimized analog wavelet circuit design reached 0.93 with an error of only 3.33%. In conclusion, it can be concluded that the research on the application of hybrid ant colony algorithm in the design of analog wavelet optimized circuits effectively improves the accuracy of wavelet approximation, and provides a new technological path for the realization of highly efficient and low-cost signal processing circuits.</div></div>","PeriodicalId":101205,"journal":{"name":"Systems and Soft Computing","volume":"7 ","pages":"Article 200184"},"PeriodicalIF":0.0,"publicationDate":"2025-01-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143148530","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Cascaded intrusion detection system using machine learning","authors":"Md. Khabir Uddin Ahamed ,&nbsp;Abdul Karim","doi":"10.1016/j.sasc.2024.200182","DOIUrl":"10.1016/j.sasc.2024.200182","url":null,"abstract":"<div><div>Cybercrime is becoming an increasing concern these days. In response to the growing cyberthreat, various intrusion detection systems have been developed and proposed to detect anomalies. However, most detection systems suffer from some common issues, such as a high number of false positives that cause regular behaviors to be detected as intrusions, as well as the system’s excessive complexity. Many single classifier models have accuracy issues since they are unable to detect certain anomalies caused by the attack’s polymorphic and zero-day behavior. The signature-based intrusion detection system (SIDS) is unable to identify zero-day intrusions. On the other side, the anomaly-based intrusion detection system (AIDS) generates a significant number of false-positive alarms. In this research, a cascaded intrusion detection system (CIDS) is proposed by combining the one-class support vector machine (OC-SVM)-based AIDS and the decision tree-based SIDS. OC-SVM is used in conjunction with the newly built Distance-Based Intrusion Classification System (DICS). SIDS that use decision trees can discover and classify anomalies. Because OC-SVM is a binary classifier, the intrusion type is determined by DICS. The suggested method aims to detect both popular and well-known zero-day attacks, as well as their type. The CIDS is evaluated using publicly available benchmark datasets, such as the Knowledge Discovery in Databases (KDD) Cup 1999 and the NSL-KDD dataset. The results of the proposed study show that CIDS outperformed both traditional SIDS and AIDS in terms of performance. Both anomalies and their types are detected with high accuracy.</div></div>","PeriodicalId":101205,"journal":{"name":"Systems and Soft Computing","volume":"7 ","pages":"Article 200182"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143148531","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Comprehensive material painting feature recognition based on spatial model","authors":"Jing Zhao ,&nbsp;Aiqin Liu","doi":"10.1016/j.sasc.2024.200181","DOIUrl":"10.1016/j.sasc.2024.200181","url":null,"abstract":"<div><div>Comprehensive material painting is an art form that uses multiple materials and techniques for creation. It combines traditional painting media with non-traditional materials, and this art form has become increasingly common in the field of contemporary art. However, due to the diversity and complexity of comprehensive material painting, traditional visual feature extraction methods are difficult to accurately identify and classify it. To address the above issues, a discriminative color space model is used to operate on the red green blue space, followed by standard processing, and finally Gabor wavelet analysis is performed on each subspace of the red green blue. The experimental results indicated that the model performed well in identification accuracy, recall, and F1 scores. Specifically, the identification accuracy of CMP-FEM reached 95.6 %, which was significantly higher than other contrast models such as IFE-MPA (85.00 %) and CR-GWFE (87.50 %). In addition, the application of the model in the field of painting restoration also showed its strong guiding ability, and the quality of the restored image was significantly improved. According to the comprehensive expert evaluation, the accuracy of the information identification was as high as 95.8 points, and the average F1 score of the repair guidance was 92.7 points, which further confirmed the practicality and accuracy of the model. These results demonstrate the superiority of the comprehensive material painting feature recognition model and provide an effective solution for the identification problem of painting authors.</div></div>","PeriodicalId":101205,"journal":{"name":"Systems and Soft Computing","volume":"7 ","pages":"Article 200181"},"PeriodicalIF":0.0,"publicationDate":"2024-12-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143148532","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Intelligent pattern design using 3D modelling technology for urban sculpture designing","authors":"Wei Wan","doi":"10.1016/j.sasc.2024.200176","DOIUrl":"10.1016/j.sasc.2024.200176","url":null,"abstract":"<div><div>3D modeling is actuality hired more and more by cities to improve urban planning and cultural protection. Sculptures in settlements are the main goal of this investigate into a novel 3D-Sculpture Architecture Estimation (3D-SAE) model. This model exploits Generative Adversarial Networks (GANs) to improve images, CNNs to extract features, and LDDNN<img>HGS-ROA, a Novel Lightweight Deep Neural Network mutual with the Hunger Games Search and Remora Optimization Method, to categorize images. The GAN-based image development module reestablishes incapacitated or low-resolution sculpture photos, and the pre-trained CNN usages transfer learning to retrieve thorough features. The LDNN, tuned via HGS and ROA, brands sculpture image classification together effective and precise. This innovative method not only improves the precision of 3D reconstruction, but it also proposals a general tool for art conservationists, urban planners, and the general public in sympathetic and taking in urban sculptures. Participating these cutting-edge tools delivers a solid basis for investigating and interpreting public art, which potentials to improve cultural asset management, art conservation, and urban planning.</div></div>","PeriodicalId":101205,"journal":{"name":"Systems and Soft Computing","volume":"7 ","pages":"Article 200176"},"PeriodicalIF":0.0,"publicationDate":"2024-12-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143148534","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Structural detection of goaf based on three-dimensional ERT technology","authors":"Nan Jia","doi":"10.1016/j.sasc.2024.200179","DOIUrl":"10.1016/j.sasc.2024.200179","url":null,"abstract":"<div><div>Goaf, as an underground space formed after mining, the accurate detection of its structure is crucial to mine safety and the stability of underground engineering. Although traditional detection methods, such as drilling and seismic methods, can provide certain information, they have limitations in terms of accuracy and economy. Therefore, this study used three-dimensional electrical resistivity tomography technology to more accurately detect the structure of the goaf due to its high resolution and non-invasive characteristics. At start, the development mechanism of the goaf was analyzed, and then the resistivity three-dimensional tomography technology was used to detect the goaf in the selected area through numerical simulation. The results show that when the surface deformation degree reaches 1.38%, the corresponding error of electrical resistivity tomography technology detection is 1.74%. When the surface deformation degree is 0.58% and 1.36% respectively, the corresponding errors of Multi-physics field monitoring method and the downhole transient electromagnetic method are 1.97% and 1.84% respectively. In the comparison of false negative rate, when the detection area reaches 76.8% of the regional detection area, electrical resistivity tomography technology has the lowest false negative rate, with a value of 2.412%. The accuracy of different methods was tested in the Jinggong and Open-pit areas. When the detection time was 0.51 s and 0.23 s respectively, the ERT method had the highest detection rate, with values approaching 98.57% and 100.00% respectively. During the whole process, the accuracy of the DTEM method was 87.85% and 99.99% respectively, which was much lower than that of the ERT method. An analysis of the low-resistivity anomaly areas in the selected study area found that the distribution of the observed areas showed uneven continuity, and its resistivity was low and significantly different from the surrounding rock formations. The above results illustrate that the main advantage of 3D ERT technology is its ability to provide real-time, high-density resistivity data, thereby enabling precise capture of subtle structural changes in the goaf. Compared with traditional methods, 3D ERT not only reduces environmental interference, but also significantly improves the efficiency of data collection and the accuracy of analysis, providing a new technical means for mine safety management and underground engineering.</div></div>","PeriodicalId":101205,"journal":{"name":"Systems and Soft Computing","volume":"7 ","pages":"Article 200179"},"PeriodicalIF":0.0,"publicationDate":"2024-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143148533","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}